Audits Claude Code session logs to identify Bash commands that required manual approval.
npx skills add https://github.com/arndvs/ctrlshft --skill press1-check-arndvs
Please help me install this Agent Skill. Skill: press1-check Source: https://github.com/arndvs/ctrlshft/tree/main/skills/press1-check Command: npx skills add https://github.com/arndvs/ctrlshft --skill press1-check-arndvs
High-intent search queries and answers about installing and using this skill.
Run the audit-permissions.py script with no arguments to scan all session logs since the last run. It prints each uncovered Bash command with a color-coded risk level and suggests allow-list rules for settings.json.
Run a permission audit to find LOW-risk commands like cat, ls, or grep that triggered prompts, then add the suggested Bash(prefix:*) rules to ~/.claude/settings.json. Re-run the audit to confirm the coverage delta.
Yes, the --auto-stop-hook mode is designed for the Claude Code Stop hook chain. It enforces a 6-hour cooldown, stays silent when nothing is actionable, and writes a pending summary when LOW-risk additions are found.
No, the script never auto-applies changes. It only surfaces findings and proposes allow rules; MEDIUM and HIGH risk commands require explicit user approval before being added to the allow-list.
The script detects JSONL files under subagents directories and tags those findings with a subagent marker. This ensures commands run by spawned subagents are included in the permission coverage analysis.
InnerScopeHearingCapture Claude Code session metrics and forward them to PostHog.
rohitg00Analyzes permission denial patterns and generates allow and deny rules for Claude Code settings.
enbiyagoralScaffold Claude Code hook scripts and register them in settings.json.
CastrozanRuns deterministic single-pass Claude Code tasks with explicit permissions.
